Postsurgical treatment with adjuvant transarterial chemoembolization in patients with hepatitis B-related hepatocellular carcinoma: A strobe-compliant article.

This study sought to develop a reliable and easy-to-use scoring model to guide the decision to perform postsurgical adjuvant transarterial chemoembolization (PA-TACE) in patients with hepatitis B-related hepatocellular carcinoma (HCC).The study included 235 consecutive patients with hepatitis B-related HCC undergoing PA-TACE at our medical center. Patients were assigned to 2 sets according to the PA-TACE date: initial (2005-2007; n = 130) and internal validation (2008-2009; n = 105) sets. With the aid of a Cox regression model, we developed a risk-scoring model from the independent predictive factors of our initial set designed as a guide for PA-TACE, and the performance of the model was validated with an internal set. External validation was also performed with an independent dataset (n = 84) to assess the discriminatory power of the scoring model.In the multivariate analysis, 4 risk factors (an increase in Child-Pugh score of at least 1 point, hepatitis B virus deoxyribonucleic acid [HBV-DNA] level >10 IU/mL, tumor diameter ≥5 cm, and the presence of vascular invasion) were significantly associated with prognosis. These factors were incorporated into a novel clinicopathological scoring model (assessment for PA-TACE [APT] risk-scoring model) ranging from 0 to 8 that was correlated with prognosis. Different survival outcomes were identified in three groups (0-2 points, 3-6 points, and 7-8 points). The risk-scoring model was further confirmed with 2 independent sets.The novel APT risk-scoring model, merging 4 prognostic factors, may achieve an optimal postsurgical prediction of PA-TACE in HBV-related HCC. The risk for an individual patient with an APT score of ≥7.0 prior to the PA-TACE, who may not profit from further PA-TACE, can be determined, and this may lead to a more appropriate choice of treatment.

Introduction

Hepatocellular carcinoma (HCC) is the most prevalent primary malignant hepatic tumor and is the 2nd-leading cause of cancer-related death.[ Hepatic resection is the most widely recommended curative treatment modality for BCLC A stage HCC. Nevertheless, the prognosis for HBV-related HCC after resection is still discouraging due to the high rate of tumor recurrence, which exceeds 70% at 5 years posthepatectomy even in patients with small tumors.[ Tumor cells are known to have aggressive spread behavior via the vasculature. Microvascular invasion (MVI) is a critical mechanism for intrahepatic metastasis and early recurrence.[ In addition, MVI displays aggressive tumor behavior and is closely correlated with larger tumor burden.[ However, there is still no recommended postsurgical treatment strategy for HCC patients with MVI, which renders the postsurgical management of these patients a major challenge.

Transarterial chemoembolization (TACE) is the treatment approach most commonly used for BCLC B stage, even for some BCLC C stage patients in China. Because the blood supply of HCC is mainly derived from the hepatic artery, injection of chemotherapeutic drugs and embolizing agents can decrease blood flow to the tumor and induce necrosis of tumor tissues at the embolization regions.[ Studies revealed that postsurgical adjuvant transarterial chemoembolization (PA-TACE) agents can improve the chance of detection and control of invisible micrometastases in the liver after resection. In China, PA-TACE was empirically used in patients with larger tumor, presence of MVI confirmed by postoperative histopathology in clinical practice. So far most research has discussed the efficacy of radical hepatectomy with or without PA-TACE for HCC patients.[ These studies indicated PA-TACE to be beneficial for patients with larger HCC, multiple nodules, and vascular invasion. Currently, few studies focused on the population associated with PA-TACE. Since large differences exist on various characteristics of patients measured before PA-TACE, this can cause significant impact on the treatment effect and prognosis of adjuvant chemotherapy after resection.

Relevant studies have shown that baseline tumor characteristics before hepatectomy have a significant impact on patient prognosis, including serum alpha-fetoprotein (AFP) and platelet count has been associated with tumor recurrence and overall survival (OS) in HCC patients.[ In addition, postsurgical pathological tumor factors such as tumor diameter, presence of vascular invasion[ are associated with the OS of HCC patients. Furthermore, postoperative persistent high viral load was associated with HCC recurrence and resulted in poor prognosis.[ Finally, as most patients with HCC have extensive liver cirrhosis, postoperative poor liver function may further worsen after adjuvant TACE and may negatively impact patient prognosis. However, the factors noted above varied to some extent due to the heterogeneity of the study populations; therefore, comprehensive predictions of survival prognosis have been difficult to make.

To the best of our knowledge, no prognostic tool currently exists to select high-risk factors associated with PA-TACE, and incorporates factors for risk stratification and prognostic prediction. The scoring model can provide evidence-based, individualized, and accurate risk estimation.[ The purpose of this study was to construct an easy-to-use risk-scoring model to help stratify patients with hepatitis B-related HCC for whom the indication for PA-TACE is optimal. The performance of the risk-scoring model was tested and verified in 2 independent sets of patients.

Methods

Patients

Consecutive patients >18 years old at the time of the hepatic resection, diagnosed with HBV-related HCC by histopathology or radiological imaging (CT/MRI scans) according to the European Association for the Study of the Liver criteria,[ and who underwent curative hepatectomy in the department of Hepatology and PA-TACE in Interventional Radiology at the Affiliated Zhongshan Hospital of Fudan University between January 2005 and December 2007 (n = 283) were screened for eligibility.

Patients with HBV-related HCC at BCLC stage A, B, or C and presurgical liver function status (Child-Pugh stage A or selected B) who received PA-TACE at 4 weeks after hepatic resection were included and formed the initial set for all further analyses.

Patients were excluded if they received liver transplantation or previous treatment for HCC. Patients who received hepatic resection despite poor liver function (Child-Pugh C) and blood tests for hepatitis B surface antigen or who were seropositive with one or more of the human immunodeficiency virus, HCV, or hepatitis D virus were also excluded.

All of the patients were rechecked in our center 4 weeks after resection. If no recurrence was found, the PA-TACE treatment strategy was recommended. If the patients were found to have single/multiple tumors during the first evaluation 4 weeks after hepatectomy, they were considered to have tumor recurrence and were excluded from this study. The diagnosis of recurrence was according to dynamic contrast-enhanced CT and/or MRI and raised serum AFP level.

A 2nd cohort of 105 patients treated in our center with PA-TACE after hepatectomy performed between January 2008 and December 2009 with the same eligibility was analyzed as an independent internal validation set. A 3rd cohort of 84 patients treated in Sun Yat-sen University Cancer Center by PA-TACE after hepatectomy between January 2006 and December 2008 with the same selection criteria was analyzed as an independent external validation set.

Collection of data

This study was approved by the Institutional Review Board of the Zhongshan Hospital and Sun Yat-sen University Cancer Center. Informed consent was obtained from all patients for their data to be used for research.

Routine presurgical imaging (chest X-ray, abdominal ultrasound, liver protocol dynamic contrast-enhanced CT and/or MRI, chest computed tomography, and bone scans) was performed 5 to 7 days before the resection.

All laboratory values, including AFP as well as liver and renal function parameters including the Child-Pugh score, were determined 1 day before the hepatectomy and 1 day before the PA-TACE session. Viral tests, including hepatitis B surface antigen, hepatitis B e antigen, and hepatitis B virus deoxyribonucleic acid (HBV-DNA) load, were performed.

Additionally, the dynamic change in the Child-Pugh score (here after referred to as Child-Pugh score increase) between the time points prehepatectomy and pre-PA-TACE was recorded. All other dynamic changes in liver/renal function-related parameters and neutrophil-to-lymphocyte ratio between prehepatectomy and pre-PA-TACE were performed as outlined in the study design and statistical analyses section.

The pathological and histological evaluations of the resected specimens, including tumor number/diameter, degree of cirrhosis, MVI, presence of capsule/infringing capsule, differentiation of tumor cells (Edmonson–Steiner classification), and presence of portal vein tumor thrombus, were recorded as well as intraoperative blood/loss transfusion and portal clamping time.

All patients received regular evaluations, including serum biochemistry, liver function test, level of AFP value, and contrast-enhanced dynamic imaging, after PA-TACE every 3 to 4 months until death or dropout from the follow-up program.

When recurrent tumor was confirmed during the study phase, the patients were actively treated with percutaneous ethanol injection, radiofrequency ablation, repeat liver resection, or TACE, according to the liver function status, tumor number, and location HCC recurrence.

Treatment procedures

Surgical procedure

Surgery was performed through a right bilateral subcostal incision. If necessary, the incision was extended to the left subcostal region. Surgeons carefully searched the abdominal cavity for the extent of local disease, extrahepatic metastases, and peritoneal seeding. The corresponding hepatic pedicle, hepatic vein, and short hepatic veins were ligated and divided. The size and number of the lesions and the relationship of the tumor to vascular structures were assessed by intrasurgical ultrasonography. Pringle maneuver was applied to occlude the blood inflow of the liver with cycles of 15 minutes clamp time/5 minutes unclamped time. Liver resection was carried out using a clamp-crushing method. Major/minor hepatic resection has been used in all surgery. Major hepatectomy was defined as resection of 4 or more liver segments. Minor hepatectomy was defined as resection of 3 or less liver segments.

PA-TACE

When the liver function of the patient had recovered at 4 weeks after resection, TACE procedure was performed for the remnant liver. Angiography of celiac, hepatic, superior mesenteric, left gastric, and bilateral inferior phrenic arteries was performed using a 4F/5F catheter to identify all feeding arteries of any obvious tumor stains in the remnant liver using the Seldinger technique. An emulsion of 2 to 10 mL of Lipiodol Ultra-Fluide (Guerbet, France) mixed with 30 to 50 mg of EADM (Pfizer, the United States of America) was then infused through a microcatheter (Progreat, TERUMO, Japan). The dosages of the chemotherapy drugs and lipiodol depended on the underlying state of liver function and body surface area.[ The criteria for liver treatment used in both institutions were similar.

Study design and statistical analyses

Numeric data are expressed as the means and SDs, and categorical data are shown as number and proportion.

The characteristics of the patients in the initial set and in the validation set are presented with descriptive statistics. OS was defined as the time from the PA-TACE until death or last follow-up. Survival curves were calculated using the Kaplan–Meier method; median OS and 95% confidence intervals are reported. Univariate analysis of the OS was performed on the estimation set. A log-rank test was performed to detect significant parameters in the univariate analysis. Variables with a strong (P ≤ 0.05) association were then entered into a stepwise Cox regression model (conditional backward selection).

Multivariate Cox regression analysis with stepwise selection was used to detect independent predictors of survival time (entry criteria for selection into the final multivariate model was P ≤ 0.05). B regression coefficients were multiplied by 3 and rounded to the nearest unit (1.00 unit) in order to obtain easy to use point numbers facilitating the bed-side calculation of the assessment for PA-TACE (APT) score.

Two validation sets (an internal validation set and an independent external validation set) were used to validate the APT scoring model. Survival curves were used to present the data. All P-values are reported using a significance level of 0.05. All calculations were performed using SPSS version 13.0 (SPSS Inc., Chicago, IL).

Results

Clinicopathological characteristics of patients

A flow chart for the initial and validation sets is shown in Fig. 1.

Figure 1

Flowchart of patient selection. HBsAg = hepatitis B surface antigen.

In the initial set (n = 130), the mean age of the patients was 53.1 years (SD, 9.8 years) and 86.9% were male. Hepatitis B infection (100%) is the most common cause of chronic liver disease, and approximately 28.5% of the patients tested positive for hepatitis B e antigen. A total of 25.4%, 59.2%, and 15.4% of patients were diagnosed as BCLC stage A, B, and C, respectively (n = 33, 77, and 20, respectively). The Child-Pugh grade prior to PA-TACE (n = 63) increased by at least 1 point compared to that before surgical resection, whereas 67 were unchanged or decreased at least 1 point. In terms of tumor factors, most patients had single tumors (60%), and the average diameter of the tumors was 6.4 cm (SD, 3.8 cm). Vascular invasion and capsular infiltration were histologically observed in 77 (59.2%) and 26 (20.0%) patients, respectively. Edmondson grade III or IV tumors were noted in 51 (39.2%) patients. Regarding operation factors, 17 (13.1%) required blood transfusion during the perioperative period. No clamping time was observed in 58 (44.6%) patients. Pathological examination revealed cirrhotic livers in most patients. HBV-DNA level reactivation (>104 IU/mL) prior to PA-TACE occurred in 53 (40.8%) patients. The median time interval between hepatectomy and PA-TACE was 35 days (range 28–40 days).

The clinical, histopathological, and surgical factors of the initial (n = 130), internal (n = 105), and external (n = 84) validation sets prior the hepatectomy and PA-TACE are summarized in Table 1. There were no significant differences in baseline characteristics between the initial and validation sets with the exception of AST (P = 0.032), Child-Pugh stage (P < 0.001), and HBV-DNA level (P = 0.031) prior to PA-TACE.

Table 1

Basal characteristics∗.

Development of the APT scoring model from the initial set

Baseline demographics were used for the univariate analysis (Table 2). Six risk factors significantly influenced prognosis: BCLC stage, tumor factor, vascular invasion, tumor diameter, HBV-DNA level, AFP level (≥200 ng/mL), and Child-Pugh score change. These 6 risk factors were used in the multivariate Cox regression analysis. After a stepwise removal of variables, 4 risk factors retained a significant predictive value for prognosis (Table 3): HBV-DNA level prior to PA-TACE, presence of vascular invasion, tumor diameter, and Child-Pugh increase by 2 points or more. BCLC stage at baseline was not significantly associated with prognosis in our population, nor was an AFP value equal to or greater than 200 ng/mL before PA-TACE. The calculated B values (regression coefficients) were multiplied by 3 and rounded to create a novel scoring model. This novel scoring model, called APT (for Assessment for Postsurgical adjuvant TACE), ranges from 0 to 8 (Fig. 3A).

Table 2

Univariate regression results for prognostic factors in the initial set.

Table 3

Multivariate stepwise backward Cox regression analysis for prognostic factors in the initial set.

Figure 3

OS in the 3 sets of patients treated by hepatectomy determined just before PA-TACE. (A) Prognostic significance according to the characteristics of the Kaplan–Meier curves with the single point scores in the initial set. (B) Initial set in accordance with the APT scoring model using a cut-off value of 3. (C) Initial set in accordance with the APT scoring model using a cut-off value of ≤2 versus 3–6 versus ≥7. (D) Internal validation set in accordance with the APT scoring model using a cut-off value of ≤2 versus 3–6 versus ≥7. (E) External set in accordance with the APT scoring model using a cut-off value of ≤2 versus 3–6 versus ≥7. APT = assessment for PA-TACE, OS = overall survival, PA-TACE = postoperative adjuvant transarterial chemoembolization.

The novel scoring model was then evaluated for the 130 patients in whom the 4 independently associated risk factors were observed. A linear decrease was observed in median OS (Fig. 2) when APT scores increased. Patients with an APT score of 0 (n = 15) had a median OS of 58 months (95% CI, 26.2–89.8). Patients with an APT score of 3 (n = 26) had a median OS of 40.2 months (95% CI, 32.2–48.2). Patients with an APT score of 5 (n = 15) had a median OS of 27.5 months (95% CI, 11.3–43.7). Patients with an APT score ranging from 7 to 8 (n = 12) had a median OS of 12.3 months (95% CI, 7.2–17.4).

Figure 2

Histograms of the evolution of median OS according to the APT scoring model in the initial set (blue bars), internal validation set (red bars), and external validation set (green bars). APT = assessment for PA-TACE, OS = overall survival, PA-TACE = postoperative adjuvant transarterial chemoembolization.

Using a threshold of 3, we were able to define 2 groups of patients with significantly different OS (P < 0.001), with a median OS of 53 months (95% CI, 38.8–67.6) for APT score ≤3 (n = 69), and a median OS of 25 months (95% CI, 19.5–30.5) for APT score >3 (n = 61) (Fig. 3B). More interestingly, using 2 different thresholds, ≤2 and≥7, we were able to enhance the prognostic value and to define three groups of patients with significantly different OS (P < 0.001). The median OS durations of the 3 groups were 61.7 months (47.7–75.7) for an APT score ≤2, 33 months (25.2–40.8) for an APT score of 3–6, and 12.3 months (7.2–17.4) for an APT score ≥7 (Fig. 3C).

Validation of the APT scoring model

For the purpose of internally validating this scoring model, using the same inclusion criteria, data were collected from an internal set of patients undergoing PA-TACE in our center between 01/2008 and 12/2009. This internal set included 105 patients (Table 1). Using the APT scoring model, scores gradually increased with decreasing OS (Fig. 2). With a cut-off value set at 3 (≤3 vs >3), the median OS durations of the 2 groups were 57.2 months (95%CI: 43.6–70.8) (n = 33) and 31.4 months (95%CI: 21.5–41.3) (n = 72) (P < 0.001), respectively. With cut-off values set at 2 (≤2) and 7 (≥7), the median OS durations of the three groups were 71.3 months (95%CI: 57.6–85) for an APT score ≤2, 35 months (95%CI: 27.4–42.6) for an APT score of 3 to 6, and 18 months (95%CI: 0–38.2) for an APT score ≥7 (P < 0.001) (Fig. 3D). Using 2 different thresholds, ≤2 and ≥7, 3 groups of patients were defined with significantly different prognoses.

For the purpose of externally validating this score, we collected data among a 3rd set of patients undergoing PA-TACE at the Cancer Center of Sun Yat-sen University. This external set comprised 84 patients (Table 1). According to the APT scoring model, OS gradually decreased with increasing score (Fig. 2). With a cut-off value set at 3 (≤3 vs >3), the median OS durations of the 2 groups were 54.6 months (95%CI: 20.6–88.6) (n = 37) and 33 months (95%CI: 20–46) (n = 47) (P = 0.003), respectively. With cut-off values set at 2 (≤2) and 7 (≥7), the median OS durations of the 3 groups were 80.1 months (95%CI: 49.8–110.5) for an APT score ≤2, 42.4 months (95%CI: 29.1–55.7) for an APT score of 3 to 6, and 18 months (95%CI: 0–21.2) for an APT score ≥7 (P < 0.001) (Fig. 3E). According to 2 different thresholds, ≤2 and ≥7, 3 groups of patients were defined with significantly different prognoses.

Discussion

Identifying the best candidates for PA-TACE is controversial because of the heterogeneity of the patients included in various studies. Additionally, the clinical factors influencing prognosis were quite diverse and have some limitations, and there is no reliable system to predict long-term prognosis. Based on our series of patients who underwent hepatectomy combined with PA-TACE for HCC, we have created a statistically predictive scoring system (APT scoring model) based on a predictive Cox regression model and tailored to the individual patient that gives accurate prognostic information in these patients. The APT scoring model is simple and easy to use, integrating 4 predictors that make up the essentials of presurgical baseline characteristics, pathological findings of the tumor, surgical operation factors, and clinical evaluation prior to PA-TACE. The predictive performance of the scoring model was further confirmed by internal and external validation sets. The scoring model provides a basis for clinicians and patients to select an appropriate therapy after radical hepatectomy for HCC (Fig. 4).

Figure 4

Proposed treatment algorithm for the postoperative adjuvant transarterial chemoembolization (PA-TACE).

Recent studies have demonstrated that the deterioration of liver function (defined as an increase in Child-Pugh score) before and after radical hepatectomy is an important risk factor for a poor prognosis. A recent study from Korea developed clinical predictive nomograms in a large cohort of HCC patients undergoing surgical resection.[ The authors found that liver function was an independent risk factor for OS identified by multivariate analyses (HR: 0.537; 95% CI, 0.434–0.665; P < 0.001), and the result was confirmed by 2 validation cohorts. Furthermore, because most patients with HCC suffer from liver cirrhosis and surgical postoperative liver function may not be fully recovered in a relatively short time, TACE may aggravate the deterioration of liver function and worsen prognosis. Sieghart et al[ analyzed the variation in the data before the 1st and 2nd TACE in 2 sets. Child-Pugh score change was considered a significant predictor of OS (HR: 4.4; 95% CI, 2.0–9.6; P < 0.001). These results are similar to those obtained in another study in France (HR: 3.03; 95% CI, 1.62–5.65; P = 0.0005).[ In keeping with the previous findings, the Child-Pugh score increase reflecting liver dysfunction is included in our proposed model.

A high HBV viral load is known to be a major risk factor for the development of HCC in patients with chronic HBV infection and for HCC recurrence after resection. Huang et al[ conducted a large comparative study of 1609 HCC patients with different serum HBV-DNA levels. They concluded that there was a significant relationship between HBV reactivation and HCC recurrence after partial hepatectomy, and a high postoperative HBV-DNA level (≥200 IU/mL) was associated with a high HCC recurrence rate. Likewise, a Taiwanese cohort study conducted by Wu et al[ confirmed that high viral loads (HBV-DNA levels >106 copies/mL) and hepatic inflammatory activity were correlated with late recurrence in hepatitis B-related HCC patients. Therefore, the HBV-DNA level is a crucial variable in our APT scoring system.

The presence of MVI is a histopathological feature that indicates aggressive behavior of the HCC, which is a powerful validated risk factor of tumor recurrence and OS following surgical treatment. Currently, the diagnosis of MVI can only reliably be determined based on the pathological histology of explanted tissue. Shim et al[ have proposed a prognostic nomogram for the prediction of recurrence and survival after HCC resection. Their results showed that MVI had high relative importance in recurrence-free survival (HR: 1.54; 95% CI, 1.21–1.95; P < 0.001) and HCC-specific survival (HR: 1.71; 95% CI, 1.26–2.31; P = 0.001) on the basis of the Cox model. Similarly, a cohort study conducted by the Singapore Medical Center[ confirmed that MVI is a strong indicator of intrahepatic metastasis in HCC, which is a better predictor of tumor recurrence and long-term prognosis following surgical resection for HCC (HR: 2.12; 95% CI, 1.52–2.97; P < 0.001). Our results are consistent with previous findings showing that including MVI in the APT scoring system is of great importance for prognoses.

Tumor diameter is a predictive covariate related to long-term prognosis in our models. Compared with patients after only hepatectomy, Sun et al[ performed a cohort study involving 322 patients to assess the effectiveness of PA-TACE for HCC patients with MVI. The maximum tumor diameter and PA-TACE were found to be independent risk factors for both recurrent-free survival and OS. However, the study did not further analyze which patient groups were suitable for PA-TACE, and it also did not build a related model based on the multivariable regression results. Our model provides a more comprehensive and powerful standard and basis for predicting the prognosis of PA-TACE in HBV-related HCC.

Our study provides new insights and guidance for determining which patients with HBV-related HCC will benefit from PA-TACE in terms of survival. Furthermore, if patients have poor liver function and/or active viral replication after hepatectomy, PA-TACE can increase liver burden and worsen liver function. In addition, the continuous high viral load after hepatectomy induces chronic inflammation in the liver remnant and may impair tumor immune surveillance, and these patients are more likely to develop multicentric carcinogenesis in the liver remnant.[ Finally, viral load is significantly associated with tumor recurrence after hepatectomy. Protecting liver function and administering antiviral treatment prior to PA-TACE not only effectively improve liver function but also decrease the chance of developing a 2nd primary HCC, which results in a better prognosis in HBV-related HCC patients.

In addition, vascular invasion and larger tumor diameter significantly affect prognosis in HCC patients. Vascular invasion shows an aggressive tumor behavior and is closely linked to large tumor burden.[ Moreover, patients with vascular invasion have a high frequency of fatal recurrence, multiple intrahepatic tumors, and extrahepatic metastasis.[ In theory, due to the rich blood supply in HCC, treatment with PA-TACE might kill or decrease the remaining tumor cells, eliminate micro-metastases to some extent, and improve long-term survival outcome. Our study suggested that vascular invasion and larger tumor diameter in HCC patients are more important and stronger risk factors for predicting prognosis, although PA-TACE could partially improve survival. Thus, single interventional treatment could not completely prevent tumor recurrence. Whether multiple preventive treatments and closer follow-up are needed, more in-depth randomized controlled trials will be required to explore and verify the basis of different BCLC stages.

The major limitation of this study is that our data were acquired retrospectively and the population was restricted to HBV-related HCC. The results may not be generalizable for prognostic prediction for patients with HCC etiology other than HBV. It will be necessary to extend our results to patients with HCC of various etiologies. Another major shortcoming is the heterogeneous time interval between hepatectomy and PA-TACE (28–40 days).

Conclusion

With the combination off our risk factors of PA-TACE, a novel, validated and widely applicable prognostic (APT) scoring model was constructed for patients with HBV-related HCC. Patients with a score of 7 or higher in the APT model prior the PA-TACE may not benefit from further PA-TACE. Our APT scoring model should be tested in prospective clinical trials.